[MPlayer-dev-eng] Re: [PATCH] Development (Was: Clean up

[Michael Niedermayer]

Hi

On Tuesday 26 February 2002 02:46, Daniel Egger wrote:
[...]
> > anyway gprof and other profilers can't show this difference...
>
> Really? You'd be surprised what gprof can show....
nothing from gprof can surprise me anymore ... i saw it output 0% for a 
function and on the next run it said 30% ...
dont understnd me wrong, gprof is a very usefull tool but i wouldnt use it 
for benchmarking, thats what "mplayer -benchmark" is for

>
> > anyway 2: gcc and other c compilers compile switch {...} to function
> > pointer array and indexes in it at the pace of your switch(). so i can't
> > see why is your version faster?
>
> Depends on the platform, the compiler, the cpu and the day of the month.
hmm, seems ur hardware is buggy if gcc outputs different binaries depending 
upon the day of the month

[...]
>
> > just imagine what would happen if everyone starts to chaneg style of
> > code...
>
> Just imagine what would happen if all core developpers agreed to a style
> and actually put it in words.
i doubt we would agree on a single style ;)

btw, when i was implementing the swscaler i looked at the source code from 
other scalers and allso on that from gimp, and imho gimps code was the most 
unreadable code i found, i mean the code from other projects like virtualdub 
which included some assembly optimized code was much more readable, allthough 
this doesnt mean that the code is unreadable to everyone ... what iam trying 
to say is that readability and style depend upon the person and i doubt that 
the gimp developers would accept it if i "fixed" it ...

[...]

Michael